Establishing or improving a simulation model of an electrolyser plant

Abstract

A method (100) for establishing and / or improving a simulation model (2) of an electrolyzer plant (1) comprising at least one electrolyzer, the simulation model (2) being configured to map one or more variables (3) characterizing an operating state (1a) of the electrolyzer plant (1) to one or more performance indicators (4) of the electrolyzer plant (1), the method comprising the steps of: predicting (110) using the simulation model (2) one or more predicted values ​​(5#) of at least one quantity (5) for a given operating state (1a) of the electrolyzer plant (1); A method (100) comprising a step (120) of obtaining one or more measured values ​​(5*) of at least one same quantity (5) related to a same given operating state (1a) of a digester plant (1), a step (130) of determining a deviation (Δ) of at least one predicted value (5#) from the respective measured value (5*), a step (140) of determining contributions and / or adjustments (2a) to the simulation model (2) from the deviations (Δ) and the simulation model (2), and a step (150) of applying the contributions and / or adjustments (2a) to the simulation model (2).

Description

[Technical field]

[0001] The present invention relates to the field of simulating the behavior of electrolyser plants with the aim of operating them in an optimal manner and detecting anomalies. [Background technology]

[0002] A large water electrolyser consists of multiple modules (AC filters, transformers, rectifiers, DC filters, one or more stacks, drying / separation tanks, module automation, (possibly) cooling) and plant level equipment (compression and storage, water treatment, buffer tanks, H2 drying and purification, transformers, plant level control and optimization).

[0003] When setting up a model of the plant behavior with the aim of operating it in an optimal way, it is not possible to deduce it from the first level of principles for several reasons: 1) The amount of information required would be too great, 2) some of the effects are not fully understood; 3) Some of the required data may not be available from the different component suppliers.

[0004] From this, the solution is to obtain behavioral information about the components from the supplier, typically based on measurements. Examples include: A typical polarization curve (U vs. I diagram) of an electrolyser stack, Efficiency values ​​for different operating points of the stack (e.g. different temperatures, different part loads, or beginning and end of life), -Efficiency curves of electrical systems, and ·Standby consumption.

[0005] From these values, parameters are calculated and used in a simplified but sufficiently accurate model of the overall system. This is somewhat similar to modeling the Earth as a solid sphere, or even as a point mass, for the purpose of making the calculations more tractable. Examples include: System efficiency (P, T, p), and - Deterioration of stack efficiency

[0006]

number

[0007] [Objective of the Invention] The aim of the present invention is to improve the accuracy of simulation models and thus to be able to draw better conclusions from these models for operating electrolyser plants.

[0008] This object is achieved by the methods set out in the independent claims. Further advantageous embodiments are detailed in the dependent claims. Summary of the Invention

[0009] The present invention provides a method for establishing and / or improving a simulation model of an electrolyser plant. The electrolyser plant comprises at least one electrolyser. The simulation model is configured to map one or more variables characterizing the operating conditions of the electrolyser plant to one or more performance indicators of the electrolyser plant. In this specification, "establishing and / or improving" should not be interpreted restrictively in the sense that the method must be able to generate a simulation model from scratch. Rather, this phrase means that contributions generated by the method, for example in the form of determined and / or adjusted model parameters, can be applied to improve an already existing model, but also when the model is being created for the first time.

[0010] In particular, a variable may comprise a quantity that can be actively set or at least measured. Examples include: the input voltage to the electrolyser plant or any part thereof (such as the electrolyser stack or electrolyser cell); the current flowing through the electrolyser plant or any part thereof; · the power input to the electrolyzer plant or any part thereof, or the rate of change of such power input; the temperature in the electrolytic cell plant or any part thereof, - Water intake pressure, the number of power cycles between higher and lower power that the electrolyzer stack or electrolyzer cell has completed, and · A set point for any low level controller in an electrolyser plant that acts on the plant or any part thereof to keep one or more quantities at or near the set point.

[0011] In particular, a performance index may comprise any quantity that is measurable, the knowledge of which in advance is desirable in the operation of an electrolyzer plant. Examples include: Hydrogen production volume, the energy efficiency of the electrolyser plant or any part thereof; the amount of waste heat generated by the electrolyser plant or any part thereof, and The deterioration state of at least one electrolyser cell or stack of the electrolyser plant.

[0012] For example, if the performance index comprises a condition index, obtaining a better prediction of this condition index earlier may enable a faster reaction to an abnormal or otherwise undesirable condition indicated by the condition index, and thus enable a faster amelioration of the abnormal or undesirable condition.

[0013] In the course of the method, for a given operating state of the electrolyzer plant, one or more predicted values ​​of at least one quantity are predicted. In particular, this predicted quantity may comprise a performance index and / or at least one physical quantity used by the simulation model for the calculation of the performance index from the variables. For example, if the simulation model first calculates a temperature and then proceeds to calculate the production of hydrogen based on this temperature, the contributions and / or adjustments of the simulation model may also be aimed at improving the accuracy with which the model predicts the temperature. This will then also improve the accuracy of the production of hydrogen calculated based at least in part on this temperature.

[0014] One or more measurements of the same at least one predicted quantity are obtained, these one or more measurements relating to the same given operating condition of the electrolyser plant.

[0015] At least one deviation of one or more predicted values ​​from each measured value is determined. For example, when predicted values ​​on the one hand and measured values ​​on the other hand for the same quantity are plotted across one or more variables, the deviation between the curves may be calculated as the sum of squared differences between the values ​​of the curves for discrete values ​​of the one or more variables, or as the integral of the area between the curves.

[0016] From this deviation and the simulation model, a contribution and / or adjustment to the simulation model is determined. For example, if the behavior of the simulation model is characterized by a set of parameters, it can be determined which parameters are the main cause of the deviation. These parameters can then be modified accordingly. The contribution and / or adjustment is applied to the simulation model. The determination of the contribution and / or adjustment can optionally be made in response to the deviation exceeding a certain threshold. That is, action on the simulation model can only be taken if the deviation is large enough.

[0017] In this way, the accuracy with which the simulation model predicts the actual behavior of the electrolyser plant can be improved. The inventors have found that there are several sources of inaccuracy when initially setting up the model.

[0018] For example, typical plant parameters on which behavioral information obtained from a component supplier is based may not perfectly match the actual parameters for which the plant was commissioned, e.g., behavioral information may have been determined by a component supplier given a water intake having a certain temperature, and at the electrolyzer plant site, the temperature of the water intake may be very different.

[0019] Additionally, component suppliers frequently determine behavior information based on a standard curve for a product series, but this standard curve may be some kind of average around which the behavior of individual pieces of hardware are scattered.

[0020] The behavioral information determined by the component manufacturer may not necessarily include all the required dependencies. For example, the efficiency of an electrolyzer cell may only be given as a function of input power, but not as a function of temperature or water pressure. Similarly, some effects may be ignored when determining the behavioral information. For example, the efficiency may only be determined for the electrolyzer stack itself, but not for the entire plant including all auxiliary equipment (referred to as Balance of Plant, BoP).

[0021] Also, some parameters on which the simulation model is based may be changed during the operation of the plant: for example, the simulation model may assume that the temperature in the electrolyzer plant is constant, but that in a particularly hot summer, the capacity of the air conditioning system may be saturated and the temperature may rise.

[0022] Finally, parameters describing degradation such as "0.1% efficiency loss per 1000 hours of operation" are quite often averages over typical values ​​and do not reflect the reality of each individual piece of equipment. This is somewhat analogous to a light bulb life specification: a 1000 hour life means that after 1000 hours, 50% of the bulbs can be expected to have failed, while the other 50% can be expected to still function.

[0023] All these uncertainties lead to a simulation model that does not reflect reality and thus may suggest set points or other operating parameters for low level controllers that are not optimal or even incorrect in the operating conditions of the electrolyzer plant. By calibrating the simulation model using actual measurements, or even by contributing to the model based on these actual measurements when creating the simulation model in the first place, the prediction errors resulting from said inaccuracies can be minimized. The predictions obtained from the simulation model are therefore a better basis for determining the optimal set points and other operating parameters.

[0024] In a particularly advantageous embodiment, the given operating conditions and measurements are obtained during normal operation of the electrolyser plant, i.e. data that arise anyway during this normal operation can be reused to refine the simulation model.

[0025] Alternatively or in combination with this, in another particularly advantageous embodiment, the electrolyzer plant is induced into a given operating state that is not part of the normal operation of the electrolyzer plant, and the response of one or more measurements to this induction is determined. In some of these special operating states, certain quantities may be determined more easily and / or with better accuracy than during normal operation of the electrolyzer plant.

[0026] For example, voltage and / or current variations may be introduced into at least one electrolyzer cell of the electrolyzer plant. The response of the current through the at least one electrolyzer cell and / or the response of the voltage across the at least one electrolyzer cell to these voltage and / or current variations may be measured. For example, the current response to the voltage variations and / or the voltage response to the current variations may be measured. This may be achieved using the existing power electronics of the electrolyzer plant. In this way, electrochemical impedance spectroscopy as known from fuel cell testing may be performed on the electrolyzer plant or any part thereof.

[0027] Alternatively, only one of the variables characterizing the operating state may be changed and at least one quantity of response to this change may be measured. In this way, an effect may be observed in isolation while other effects are excluded. This does not introduce the additional imprecision of having to disentangle different effects from one another.

[0028] In particular, the temperature of at least one electrolyser cell or stack of the electrolyser plant is chosen as the variable to be varied, as temperature is one of the parameters that can vary most significantly between the creation of the original model on the one hand and the productive operation of the electrolyser plant on the other hand.

[0029] In a particularly advantageous embodiment, the contributions and / or adjustments to the simulation model include: Changes to parameters in parameterized functions of variables; Changes to the set of variables, changing the functions of the variables to functions of different orders and / or types and / or to different functions; Changes to the level of complexity of the simulation model; Changes to the set of physical effects considered or ignored by the simulation model; It has one or more of the following:

[0030] For example, the simulation model may calculate the predicted quantity as a polynomial of a given degree in one or more variables, such as a linear or multilinear function of one or more variables. The coefficients of this polynomial may then be adapted according to the deviations. However, the degree of the polynomial may also be increased, for example so that a previously linear function becomes a quadratic function. The inverse problem of obtaining the contributions and / or adjustments may be solved using any suitable known mathematical method.

[0031] It is also generally possible to change the level of complexity of the simulation model depending on the determined deviations. For example, the simulation model may initially start as a relatively simple model, and if this proves not to be accurate enough, it may be expanded to a more complex model that is computationally more expensive. In this way, the higher computational expense is only used when it is actually needed, and is not "wasted" in obvious situations that can be handled equally well by a simple model.

[0032] For example, determining the contributions and / or adjustments may comprise fitting at least one function of the variables and / or calculations derived from the variables to the measurements. The contributions and / or adjustments may then be optimized to achieve the best match between the fitted function and the measurements according to any suitable criterion, such as least squares error.

[0033] In a further particularly advantageous embodiment, during fitting, the measurements are How often they occur, and their duration, - Their accuracy estimates and are weighted according to one or more of the following:

[0034] For example, if measurements are taken during normal operation of the electrolyser plant and the electrolyser plant is operated at a particular operating point (such as a certain power level) most of the time, then most of the taken measurements will be concentrated around that operating point, i.e. the frequency of measurements will be highest there. Two conclusions can be drawn from this. First, from a statistical point of view, the true value to which the model should match can be determined very accurately from a lot of taken measurements related to the primary operating point of the electrolyser plant. Second, if the electrolyser is operated at the primary operating point most of the time, it is most important at this primary operating point to have a good agreement between the simulation model and the measurements.

[0035] Weighting the measurements according to their age is particularly advantageous when drift occurs over time, for example when measurements relate to some entity that is subject to degradation, older measurements should have less influence on the contribution and / or adjustment than more recent measurements.

[0036] Another example where weighted measurements are advantageous is when information about the measurement error is available. Typically, the accuracy of hydrogen mass flow measurements depends on the mass flow rate, so the higher the mass flow rate, the higher the measurement error. The measurement weight in this situation is given by the inverse of the measurement error.

[0037] In a further particularly advantageous embodiment, the evolution of the contributions and / or adjustments to the simulation model over time is tracked, from which degradation behavior and / or abnormal operating states of the electrolyzer plant can be evaluated, in this way appropriate reactions can be initiated at an earlier time.

[0038] The method may be computer-implemented in whole or in part. The present invention therefore also relates to one or more computer programs having machine-readable instructions that, when executed on one or more computers and / or computing instances, cause the one or more computers to perform the methods described above. In this context, virtualization platforms, hardware controllers, network infrastructure devices (such as switches, bridges, routers, or wireless access points), and end devices in a network (such as sensors, actuators, or other industrial field devices) capable of executing machine-readable instructions should also be considered as computers.

[0039] The present invention therefore also relates to a non-transitory storage medium and / or a download product comprising one or more computer programs. A download product is a product that can be sold in an online shop for immediate fulfillment by download. The present invention also provides one or more computers and / or computing instances comprising one or more computer programs and / or one or more non-transitory machine-readable storage media and / or a download product.

[0040] In the following the invention is illustrated using figures, without any intention of limiting the scope of the invention. [Brief description of the drawings]

[0041] [Figure 1] 1 is an illustrative embodiment of a method 100 for establishing and / or improving a simulation model 2 of an electrolyzer plant 1 . [Diagram 2] 2 are two possible illustrative adjustments to the original simulation model 2 based on the same set of measurements. [Diagram 3] Illustrative adjustments to the original simulation model 2 based on a weighted consideration of the measurements. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0042] 1 is a schematic flow chart of an embodiment of a method 100 for establishing and / or improving a simulation model 2 of an electrolyzer plant 1. The simulation model 2 is configured to map one or more variables 3 characterizing an operating state 1a of the electrolyzer plant 1 to one or more performance indicators 4 of the electrolyzer plant 1.

[0043] In step 110, starting from a given operating state 1a of the electrolyser plant 1, the simulation model 2 is used to predict one or more predicted values ​​5# of at least one determined quantity 5.

[0044] According to block 111 , this determined quantity 5 may comprise a performance index 4 and / or at least one physical quantity used by the simulation model 2 for the calculation of the performance index 4 from the variables 3 .

[0045] In step 120, one or more measured values ​​5* of the same at least one determined quantity 5 are obtained. The one or more measured values ​​5* relate to a given operating state 1a of the electrolyser plant 1 which is the same as the predicted value 5#.

[0046] According to block 121, the given operating state 1a and the measured values ​​5* may be obtained during normal operation of the electrolyser plant 1. That is to say, the one or more variables 3 characterizing the operating state 1a and the measured values ​​5* of the sought quantities 5 which may ultimately give rise to the performance index 4 may be obtained from data already available in the course of the normal operation. The normal operation is not changed.

[0047] According to block 122, the electrolyzer plant 1 can be induced into a given operating state 1a that is not part of the normal operation of the electrolyzer plant 1. According to block 123, the response of one or more measurements 5* to this induction can then be determined.

[0048] In one illustrative embodiment, according to block 122a, voltage and / or current variations may be introduced into at least one electrolyzer cell of the electrolyzer plant 1. According to block 123a, the response of the current through the at least one electrolyzer cell and / or the response of the voltage across the at least one electrolyzer cell to these voltage and / or current variations may then be measured.

[0049] In another illustrative embodiment, according to block 122b, only one of the multiple variables 3 characterizing the operating state 1a may be changed. The response of the at least one quantity 5 to this change may then be measured.

[0050] In step 130, at least one deviation Δ of one or more predicted values ​​5# from a respective measured value 5* is determined.

[0051] In step 140, from the deviation Δ and the simulation model 2, a contribution and / or adjustment 2a to the simulation model 2 is determined.

[0052] According to block 141, the contribution and / or adjustment 2a is A change to a parameter in a parameterized function of variable 3, Changes to the set of variable 3, changing the function of variable 3 to a function of a different order and / or type and / or to a different function; Changes to the simulation model's complexity level of 2 and Changes to the set of physical effects taken into account or ignored by the simulation model 2; may comprise one or more of:

[0053] According to block 142, determining 140 the contributions and / or adjustments 2a may comprise fitting at least one function of variables 3 and / or calculations derived from variables 3 to the plurality of measurements 5*.

[0054] According to block 142a, during fitting 142, measurement 5* is How often they occur, and their duration, - Their accuracy estimates and may be weighted according to one or more of:

[0055] In step 150, the contributions and / or adjustments 2a are applied to the simulation model 2.

[0056] In step 160, the evolution 2a' of the contributions and / or adjustments 2a to the simulation model 2 can be tracked over time. From this evolution 2a', in step 170, the degradation behavior 1b and / or abnormal operating conditions 1c of the electrolyzer plant 1 can then be evaluated. This can be used in any suitable way to the benefit of the electrolyzer plant 1. For example, the electrolyzer plant 1 can be operated in a way that is likely to return the electrolyzer plant 1 from the abnormal conditions 1c to normal operating conditions 1a, or at least slow down (possibly reverse) the degradation behavior 1b. Alternatively or in combination with this, the degradation behavior 1b, the abnormal operating conditions 1c, and / or even the evolution 2a' of the contributions and / or adjustments 2a themselves can be graphically rendered for an operator of the electrolyzer plant 1, for example on a graphical dashboard. This provides the operator with a quick overview on the health state of the plant components.

[0057] In Figure 2, measured values ​​5* of the hydrogen (H2) output rate of an exemplary electrolyser plant 1 as performance indicators 4 and desired quantities 5 are plotted over the power input to the electrolyser plant 1 as variable 3. In the same figure, predicted values ​​5# of the desired quantities 5 obtained from different simulation models are plotted.

[0058] In the example shown in Figure 2, the original simulation model 2 is a simple linear model where hydrogen output is strictly proportional to power input. This original model appears to match the measurements 5* for very low and very high power inputs, but not for intermediate power inputs.

[0059] By adding an offset as adjustment 2a to the strictly proportional original model 2, a better agreement with the measurements 5* is obtained on average. However, this comes at the cost of a worse agreement for very low and very high power inputs. If the electrolyzer plant 1 is operated at a higher power input most of the time, it is better to stick to the original simulation model 2.

[0060] Another illustrative adjustment 2a* goes beyond simply changing the coefficients of the linear model. Rather, the dependence of hydrogen output on power input is changed from linear to quadratic. This results in a much better agreement with the measurements 5*. In particular, it seems more physically plausible that hydrogen production only starts when power input exceeds a certain threshold. For this reason, in this example, it is better to allow the change from a linear to a quadratic function.

[0061] Figure 3 also shows the measured values ​​5* of hydrogen output rate as performance indicator 4 and the desired quantity 5 plotted across power input as variable 3. Compared to Figure 2, this scenario differs in that the measured values ​​5* are concentrated at higher power inputs. As a result, the measured values ​​5* for higher power inputs are given a higher weight when determining adjustments 2a to model 2.

[0062] In the example shown in Figure 3, sticking to the linear dependence of the hydrogen output on the power input leads to a better agreement with the measured value 5* for high power input than switching to a quadratic dependence. On the other hand, the physical implausibility of a non-zero hydrogen output even at zero power input reappears. However, for the application at hand, this may not be as important as a higher accuracy in the regime of high power input.

[0063] [List of reference numbers] 1. Electrolyser Plant 1a Operation state of electrolytic cell plant 1 1b Deterioration behavior of electrolytic cell plant 1 1c Abnormal operating condition of electrolytic cell plant 1 2 Simulation model of electrolytic cell plant 1 2a, 2a* Contribution and / or adjustment of simulation model 2 2a' Evolution of contribution 2a in time 3. Variables characterizing operating state 1a 4 Performance index for electrolytic cell plant 1 5. Required quantity 5* Measurement value of desired quantity 5 5# Estimated value of quantity 5 100 Method for establishing and / or improving a simulation model 2 110 Predicting the predicted value 5# of the desired quantity 5 111 Specific choices for the quantity sought 5 120 Obtaining a measurement value 5* of the desired quantity 5 121 Obtain state 1a, measurement 5* during normal operation 122 Inducing the plant 1 into a special operating state 1a 122a Introducing voltage and / or current variations 122b Changing only one of several variables3 Determine the response of measure 5* to 123 leads 123a Determining response to voltage and / or current variations 123b Determining the response to a change in a single variable 130 Determining deviation Δ 140 Determining the contribution and / or adjustment 2a 141 Specific choices regarding contribution and / or adjustment 2a 142 Fitting a function of variable 3 to measurements 5* 142a Weighting measurements5* during fitting 150 Applying contributions and / or adjustments 2a to simulation model 2 160 Tracking the progress 2a' over time 170 Evaluating the deterioration behavior 1b and abnormal state 1c from the progress 2a' Δ Deviation of predicted value 5# from measured value 5*

Claims

1. A method (100) for establishing and / or improving a simulation model (2) of an electrolyser plant (1) comprising at least one electrolyser, said simulation model (2) being configured to map one or more variables (3) characterizing an operating state (1a) of said electrolyser plant (1) to one or more performance indicators (4) of said electrolyser plant (1), said method comprising the steps of: - predicting (110) one or more predicted values ​​(5#) for at least one quantity (5) for a given operating state (1a) of the electrolyser plant (1) using said simulation model (2); - obtaining (120) one or more measurements (5*) of the same at least one quantity (5) related to the same given operating state (1a) of said electrolyser plant (1); - determining (130) at least one deviation (Δ) of one or more of said predicted values ​​(5#) from a respective said measured value (5*); - determining (140) from said deviation (Δ) and said simulation model (2) a contribution and / or adjustment (2a) to said simulation model (2); - applying (150) said contributions and / or adjustments (2a) to said simulation model (2); A method (100).

2. 2. The method (100) of claim 1, wherein the given operating conditions (1a) and the measurements (5*) are obtained (121) during normal operation of the electrolyser plant (1).

3. 3. The method (100) according to claim 1 or 2, wherein the electrolyser plant (1) is induced (122) into a given operating state (1 a) that is not part of normal operation of the electrolyser plant (1) and a response (123) of one or more measurements (5*) to this induction is determined.

4. 4. The method (100) of claim 3, wherein voltage and / or current variations are introduced (122a) into at least one electrolyzer cell of the electrolyzer plant (1) and a response of a current through and / or a voltage across at least one of the electrolyzer cells to these voltage and / or current variations is measured (123a).

5. 5. The method (100) according to claim 3 or 4, wherein only one of a plurality of variables (3) characterizing said operating state (1a) is changed (122b) and the response of at least one of said quantities (5) to this change is measured (123b).

6. 6. The method (100) of claim 5, wherein the temperature of at least one electrolyzer cell of the electrolyzer plant (1) is selected as the variable to be varied.

7. 7. The method (100) according to any one of the preceding claims, wherein the performance indicators (4) comprise the production amount of hydrogen, and / or the health state of at least one electrolyser cell of the electrolyser plant (1), and / or the energy efficiency of the electrolyser plant (1), and / or the amount of waste heat generated by the electrolyser plant (1).

8. The method (100) according to any one of claims 1 to 7, wherein the quantities (5) comprise (111) the performance index (4) and / or at least one physical quantity used by the simulation model (2) for the calculation of the performance index (4) from the variables (3).

9. The contributions and / or adjustments (2a) to the simulation model (2) may include: - changes to parameters in a parameterized function of the variable (3); - a change to the set of variables (3); - changing the function of the variable (3) to a function of a different order and / or type and / or to a different function; - changes to the level of complexity of the simulation model (2); - modifications to the set of physical effects taken into account or ignored by the simulation model (2); The method (100) of any one of claims 1 to 8, comprising one or more of the following (141).

10. The method (100) according to any one of claims 1 to 9, wherein determining (140) the contribution and / or adjustment (2a) comprises fitting (142) at least one function of the variables (3) and / or of a calculation derived from the variables (3) to a plurality of measurements (5*).

11. During the fitting (142), the measurements (5*) are How often they occur, and - Their elapsed time, - their accuracy estimates and The method (100) of claim 10, wherein the weighting (142a) is based on one or more of:

12. Tracking (160) the evolution (2a') of the contributions and / or adjustments (2a) to the simulation model (2) over time; evaluating (170) the degradation behavior (1b) and / or abnormal operating conditions (1c) of the electrolyzer plant (1) from said evolution (2a'); The method (100) of any one of claims 1 to 11, further comprising:

13. A computer program comprising machine-readable instructions which, when executed on one or more computers, cause said one or more computers to perform the method of any one of claims 1 to 12.

14. A non-transitory storage medium and / or download product comprising a computer program according to claim 13.

15. One or more computers comprising a computer program according to claim 13 and / or a non-transitory storage medium and / or a download product according to claim 14.

Images